|Home | About | Journals | Submit | Contact Us | Français|
Apoptosis-induced proliferation (AiP) is a compensatory mechanism to maintain tissue size and morphology following unexpected cell loss during normal development, and may also be a contributing factor to cancer and drug resistance. In apoptotic cells, caspase-initiated signaling cascades lead to the downstream production of mitogenic factors and the proliferation of neighbouring surviving cells. In epithelial cells of Drosophila imaginal discs, the Caspase-9 ortholog Dronc drives AiP via activation of Jun N-terminal kinase (JNK); however, the specific mechanisms of JNK activation remain unknown. Here, we show that caspase-induced activation of JNK during AiP depends on an inflammatory response. This is mediated by extracellular reactive oxygen species (ROS) generated by the NADPH oxidase Duox in epithelial disc cells. Extracellular ROS activate Drosophila macrophages (hemocytes), which in turn trigger JNK activity in epithelial cells by signaling through the TNF ortholog Eiger. We propose that in an immortalized (‘undead’) model of AiP, signaling back and forth between epithelial disc cells and hemocytes by extracellular ROS and TNF/Eiger drives overgrowth of the disc epithelium. These data illustrate a bidirectional cell/cell communication pathway with implication for tissue repair, regeneration and cancer.
Following significant apoptotic cell death, apoptosis-induced proliferation (AiP) is a form of compensatory proliferation that can regenerate lost tissue via additional or accelerated cell divisions and is defined as the process by which apoptotic cells actively stimulate surviving cells to divide . In Drosophila, there is mounting evidence that AiP is driven by mitogenic signals produced by apoptotic caspases in the dying cell [2–8]. Likewise, numerous studies in several regenerative model organisms including Hydra, planarians, newt, zebrafish, Xenopus and mammals have validated the concept of caspase-driven AiP (reviewed in [9, 10]). Furthermore, AiP may also be a contributing factor for the development of cancer and drug resistance [11–14].
Caspases are highly specific cell death proteases. In apoptotic cells, activated initiator caspases such as Caspase-9 and its Drosophila ortholog Dronc cleave and activate effector caspases such as Caspase-3 and its orthologs DrICE and Dcp-1 which trigger apoptosis (reviewed in [9, 10]). In addition to activating effector caspases, Dronc can also promote AiP through activation of Jun N-terminal kinase (JNK) signaling [4, 5, 15–18]. However, the specific mechanisms by which Dronc activates JNK are not known. Therefore, to facilitate screening for genes and mechanisms involved in AiP, we have developed the ey>hid-p35 model in Drosophila . In this AiP model, the pro-apoptotic gene hid and the caspase inhibitor p35 are co-expressed under control of the eyeless-Gal4 (ey-Gal4) driver in the anterior eye imaginal disc. Because P35 very specifically inhibits the effector caspases DrICE and Dcp-1 in Drosophila , hid/p35-expressing cells initiate the apoptotic process by activating Dronc, but cannot execute apoptosis due to effector caspase inhibition by P35, thus producing “undead” cells [2–5, 15]. Furthermore, because undead cells do not die, continued Dronc signaling initiates a feedback amplification loop that involves Dronc, JNK and Hid which further amplifies the signals for AiP [16, 20]. Thus, Dronc chronically signals for AiP and triggers overgrown imaginal discs which in the case of ey>hid-p35 produces overgrowth of adult heads with pattern duplications compared to control (ey>p35) animals (Figure 1A,B) . In ey>hid-p35 eye imaginal discs, the anterior part of the eye disc where ey-Gal4 is expressed is overgrown at the expense of the posterior eye field . This reduction of the posterior eye field can be visualized using the photoreceptor marker ELAV (Figure S1A,B). We are using the normalization of the ELAV pattern in the posterior eye field in various genetic backgrounds as indicator of the suppression of ey>hid-p35–induced overgrowth at the disc level (Figure S1C).
Here, using the ey>hid-p35 model of AiP to investigate the mechanisms by which Dronc activates JNK signaling, we show that Dronc activity in epithelial disc cells promotes activation of the NADPH oxidase Duox which generates extracellular reactive oxygen species (ROS). Extracellular ROS activate hemocytes, Drosophila macrophages, at undead tissue. Activated hemocytes in turn release the TNF ligand Eiger which promotes JNK activation in epithelial cells and promotes AiP. These data illustrate a bidirectional cell/cell communication pathway with implication for tissue repair, regeneration and cancer.
When reactive oxygen species (ROS) accumulate indiscriminately within cells, they can be toxic leading to oxidative stress and possible cell death. However, at lower, controlled levels, ROS can have specific roles in growth control, proliferation and differentiation . Recent studies have demonstrated critical requirements for ROS during wound healing and regeneration, and in certain contexts via activation of JNK [22–24]. In order to examine the role of ROS in AiP, we assessed in vivo ROS levels in Drosophila imaginal discs using the ROS-reactive dyes dihydroethidium (DHE) and the fluorescein based H2-DCF-DA . In undead eye imaginal discs, ROS are dramatically increased compared to control discs (Figure 1D,E,H,I). This increased ROS production in undead tissue is dependent on Dronc activity (Figure 1F,J) consistent with the suppression of the adult head overgrowth phenotype (Figure 1C) and the normalization of the ELAV pattern by dronc mutations (Figure S1C). We also detected increased ROS in undead wing imaginal discs (nub>hid-p35) (Figure 1L,M) suggesting that the production of ROS in response to caspase activation is not tissue-specific. These data imply that ROS can be generated in developing epithelial tissues following initiator caspase activation, independently of cell death execution.
We also examined if ROS are produced in a p35-independent AiP model in which apoptosis is temporally induced and cells are allowed to die [5, 17, 18, 26]. In this AiP model, hid expression is induced for 12h in the dorsal half of the eye disc using dorsal eye-(DE-)Gal4 (DEts>hid) during 2nd or 3rd larval instar . ROS are observed immediately after apoptosis induction (Figure 1N,O; Figure S2A,D) suggesting that its production is an early event in the death and regeneration process. In this AiP model, ROS levels can be detected for up to 24 hours after apoptosis induction (Figure S2B,C). These data are consistent with a recent report about ROS production in a p35-independent model in wing imaginal discs .
To determine if there is a functional requirement for ROS in AiP, we mis-expressed ROS-reducing enzymes in the undead AiP model. However, mis-expression of cytosolic SOD and cytosolic catalase transgenes did not significantly suppress ey>hid-p35-induced overgrowth (Figure 2A,E). In contrast, mis-expression of two extracellular catalases, immune-regulated catalase (IRC) and a secreted human catalase (hCatS), does suppress ey>hid-p35-induced overgrowth (Figure 2B,E; see detailed statistical analysis in Figure S3F). Consistently, mis-expression of hCatS results in a strong reduction of ROS in undead eye discs (Figure S3A–C). These observations suggest that extracellular ROS are required for AiP following induction of apoptosis.
Two enzymes known to generate extracellular ROS are the transmembrane NADPH oxidases Nox and Duox . To examine if either of these enzymes are involved in ROS production during AiP, we knocked down their expression by RNAi. Targeting Nox did not significantly suppress the AiP overgrowth phenotype in ey>hid-p35 animals (Figure 2C,E; see summary in Figure S3E). In contrast, RNAi against Duox produced a suppression of the AiP overgrowth phenotype (Figure 2D,E; Figure S3D,E). Three independent RNAi transgenes for both of these genes gave consistent results (Figure S3D,E).
Because ey>hid-p35 imaginal discs cause overgrowth due to increased cell proliferation (Figure 2F) , we examined if the suppression of overgrowth by Duox RNAi and hCatS expression is due to reduction of mitotic activity in ey>hid-p35 discs by PH3 labeling. This was indeed found to be the case (Figure 2G″,H″,I″; quantified in Figure 2F). Concomitantly, a normalization of the ELAV pattern was observed in ey>hid-p35 discs after reduction of ROS by Duox RNAi and hCatS expression (Figure 2G–I).
As Duox was disc-autonomously inhibited in undead cells (using ey-Gal4), these data indicate that extracellular ROS originate from the same cells that have activated Dronc, consistent with the observation that ROS production requires Dronc (Figure 1F,G,J,K). Combined, Duox activity in undead cells produces extracellular ROS, which is required for AiP-induced overgrowth.
Next, we examined the position of ROS function relative to JNK, which is a critical mediator of AiP [4, 5, 17, 18]. Using MMP1 as a marker of JNK activity , we observed that JNK activity is strongly reduced in undead eye discs that overexpress hCatS or Duox RNAi (Figure 2G′–I′). Furthermore, activation of JNK by overexpression of a constitutively active JNKK (hepCA) in a dronc−/− background does not result in the generation of ROS (Figure S4). These data are consistent with a model in which extracellular ROS are produced and acting upstream of JNK. Because JNK markers and cleaved Caspase 3 labelings overlap in undead cells [4, 5, 18], we propose that extracellular ROS signal back to undead cells – directly or indirectly - to turn on JNK, and thus activate downstream mitogen production for AiP.
To identify the function of extracellular ROS, we considered that they may activate hemocytes, Drosophila innate immune cells, as has been observed in Drosophila embryos . Indeed, the pan-hemocyte anti-Hemese (H2) antibody  revealed that hemocytes are attached to both control and undead eye imaginal discs (Figure 3A,B). In wing imaginal discs, where no hemocytes are attached to control discs, there is a strong increase in the number of hemocytes attached to undead tissue (Figure S5A,B). There are three different types of hemocytes in the Drosophila larva: plamatocytes, lamellocytes and crystal cells . Cell type-specific markers identify the hemocytes attached to undead tissue as macrophage-like plasmatocytes (Figure S5C–E).
Most strikingly, although hemocytes are present at control eye imaginal discs, they change morphology and location when they are attached to undead tissue. At control eye discs, they tend to form large cell clusters (Figure 3A,C) which are located around the border between anterior proliferating tissue and the posterior differentiating photoreceptors. In contrast, hemocytes on undead discs are enlarged and often present as single cells, are less spherical and extent cellular protrusions (Figure 3B,D) which may assist in signaling between hemocytes and undead epithelial cells . Furthermore, they also attach to undead tissue that displaces part of the posterior eye tissue as visualized by disrupted ELAV labeling (compare Figure 3B,B′ with 3A,A′). A similar morphology of activated hemocytes was observed in undead wing imaginal discs (Figure S5B). Upon loss of ROS by expression of the extracellular catalase hCatS or Duox RNAi, hemocyte recruitment is strongly impaired and ELAV labeling is normalized (Figure 3E,F). These results suggest that extracellular ROS activate hemocytes at undead tissue.
Interestingly, we also noted that the labeling of ey>hid-p35 discs with cleaved caspase-3 antibody (CC3) is reduced, but not completely absent, when ROS are removed by Duox RNAi and hCatS expression (compare Figure 3E″,F″ with 3B″). Because CC3 is a marker of active Dronc  which is activated by ey-driven hid, and because CC3 labeling is not completely absent, the reduction of CC3 labeling after removal of extracellular ROS suggest that ROS participate in the feedback amplification loop between Dronc, JNK and Hid that has previously been described during stress-induced apoptosis [16, 20].
Next, we investigated whether hemocytes promote or restrict the overgrowth of undead tissue. To address this question, we analyzed ey>hid-p35 animals that are mutant for serpent (srp), a GATA-type Zn-finger transcription factor that is required for hemocyte differentiation . Of particular importance is the srpneo45 allele that specifically affects srp’s requirement for hemocyte differentiation, but leaves other functions of srp intact . The adult overgrowth phenotype of ey>hid-p35 animals is dramatically suppressed when heterozygous mutant for srpneo45 (Figure 4A,B,E; see detailed statistics in Figure S3G). Similar results were obtained for a different srp allele, srp01549 (Figure 4C,E; Figure S3G). In contrast, inactivation of srp specifically in the eye imaginal disc epithelium using three independent RNAi lines does not result in suppression of the overgrowth phenotype (Figure 4D,E; Figure S3G) indicating that srp is required non-disc autonomously for AiP. Given that srpneo45 specifically affects hemocyte function, we conclude that hemocytes are required for AiP in the ey>hid-p35 model.
Concomitantly, loss of srp function reduces JNK activity in undead tissue (Figure 4F,G,H). To exclude the possibility that JNK activity is responsible for hemocyte recruitment to undead disc tissue, we activated JNK by expression of hepCA. This experiment was done in a dronc−/− background to block the feedback amplification loop which may otherwise activate hemocytes. However, under these conditions, hemocytes are not activated (Figure S6) further substantiating that hemocytes are acting upstream of JNK. These results suggest that undead tissue produces extracellular ROS through activation of Duox, which triggers an inflammatory response by activating hemocytes. Hemocytes in turn are required for JNK activation and the overgrowth of the undead epithelial disc tissue.
Interestingly, ey>hid-35 eye discs mutant for srp also show a reduction, but not complete absence, of caspase activity as detected by CC3 (Figure 4I,J,K). This observation suggests that hemocytes participate in the feedback amplification loop in apoptotic cells described previously [16, 20].
We frequently observed that hemocytes and epithelial cells expressing the JNK marker puc-lacZ are in direct contact (Figure 5A), further confirming the notion that hemocytes promote JNK activation in epithelial disc cells. This observation suggests that hemocytes release one or more signals that induce JNK activation in disc cells and promote AiP.
One signal known to induce the JNK cascade in epithelial cells is Eiger, a TNF-like ligand in Drosophila [37–39]. To test a requirement of Eiger for AiP, we generated ey>hid-p35 flies in homozygous eiger mutant background. Under these conditions, the overgrowth of the adult head is strongly suppressed (Figure 5B,C,E; see detailed statistical analysis in Figure S3H). Loss of eiger also results in loss of JNK activity in ey>hid-p35 discs (Figure 5F,G,G′,H,H′) consistent with a role of Eiger signaling for JNK activation [37–39].
In contrast, RNAi targeting eiger specifically in the eye disc (using ey-Gal4), does not suppress the overgrowth phenotype  (Figure 5E; Figure S3H) suggesting a non-disc autonomous function of Eiger. A putative role of Eiger in hemocytes is supported by immunofluorescent analysis using anti-Eiger antibody. In contrast to ey>p35 controls, we observed Eiger protein in hemocytes attached to undead epithelial tissue (Figure 6; arrows in D and D′). There is also increased diffused Eiger labeling immediately outside of hemocytes at undead disc tissue (arrowhead in Figure 6D′,D″). These data suggest that Eiger is upregulated in and secreted by hemocytes attached to undead tissue consistent with work by another group which showed increased Eiger protein in hemocytes in a different context .
There are two putative Eiger/TNF receptors encoded in the Drosophila genome, wengen and grindelwald (grnd) [39, 41, 42]. We have previously shown that wengen is not required disc-autonomously for overgrowth of undead tissue  (see also Figure 5E; Figure S3H). In contrast, RNAi targeting grnd in the disc strongly suppressed the overgrowth of ey>hid-p35 animals (Figure 5D,E; Figure S3H) suggesting a disc-autonomous requirement of grnd. Consistently, Grnd uses the same regulatory factors for JNK activation, Traf2 and Tak1 , that are also required for JNK activation in AiP . Additionally, upon grnd knockdown, the ectopic JNK activity in the undead tissue is lost (Figure 5I). These results provide evidence that Eiger/Grnd-dependent activation of JNK serves as intermediary of hemocyte/disc crosstalk, which is required for AiP and overgrowth of undead epithelial tissue.
The role of ROS as a regulated form of redox signaling in damage detection and damage response is becoming increasingly clear . Here, we have shown that in Drosophila extracellular ROS generated by the NADPH oxidase Duox drive compensatory proliferation and overgrowth following hid-induced activation of the initiator caspase Dronc in developing epithelial tissues (Figure 7). We find that at least one consequence of ROS production is the activation of hemocytes at undead epithelial disc tissue. Furthermore, our work implies that extracellular ROS and hemocytes are part of the feedback amplification loop between Hid, Dronc and JNK (Figure 7) that occurs during stress-induced apoptosis [16, 20]. Finally, hemocytes release the TNF ligand Eiger which promotes JNK activation in epithelial disc cells (Figure 7).
This work helps to understand why JNK activation occurs mostly in apoptotic/undead cells, but occasionally also in neighbouring surviving cells [4, 5, 17, 18]. Because our data indicate that hemocytes trigger JNK activation in epithelial cells, the location of hemocytes on the imaginal discs determines which epithelial cells receive the signal for JNK activation (Figure 7). Nevertheless, we do not exclude the possibility that there is also an autonomous manner of Dronc-induced JNK activation in undead/apoptotic cells as indicated in Figure 7.
In the context of apoptosis, hemocytes engulf and degrade dying cells . However, there is no evidence that hemocytes have this role in the undead AiP model. We do not observe CC3 material in hemocytes attached to undead tissue. Therefore, the role of hemocytes in driving proliferation is less clear and likely context-dependent. In Drosophila embryos, hemocytes are required for epidermal wound healing, but this is a non-proliferative process . With respect to tumor models in Drosophila, much of the research to date has focused on the tumor suppressing role of hemocytes and the innate immune response [44–46]. However, a few reports have implicated hemocytes as tumor-promoters in a neoplastic tumor model [40, 47]. Consistently, in our undead model of AiP, we find that hemocytes have an overgrowth- and tumor-promoting role. Therefore, the state of the damaged tissue and the signals produced by the epithelium may have differential effects on hemocyte response.
In a recent study, ROS were found to be required for tissue repair of wing imaginal discs in a regenerative (p35-independent) model of AiP , consistent with our work. While a role of hemocytes was not investigated in this study , it should be noted that p35-independent AiP models do not cause overgrowth, while undead ones like the ey>hid-p35 AiP model do. It is therefore possible that ROS in p35-independent AiP models are necessary for tissue repair independently of hemocytes, while ROS in conjunction with ROS-activated hemocytes in undead models mediate the overgrowth of the affected tissue. Future work will clarify the overgrowth-promoting function of hemocytes. These considerations are reminiscent of mammalian systems, where many solid tumors are known to host alternatively activated (M2) tumor-associated macrophages, which promote tumor growth and are associated with a poor prognosis (reviewed in ).
As tumors are considered “wounds that do not heal” , we see the undead model of AiP as a tool to probe the dynamic interactions and intercellular signaling events that occur in the chronic wound microenvironment. Future studies will investigate the specific mechanisms of hemocyte-induced growth and the tumor promoting role of inflammation in Drosophila as well as roles of additional tissue types, such as the fat body, on modulating tumorous growth.
See Supplemental Information for experimental procedures.
We would like to thank István Andó, Eric Baehrecke, Marc Freeman, Won Jae Lee, Masayuki Miura, Neal Silvermann, the Bloomington Drosophila Stock Center, the Vienna Drosophila Resource Center, and the Developmental Studies Hybridoma Bank for fly lines and reagents; Ernesto Perez for input and technical expertise; Latisha Eilijo for technical assistance. CEF would like to thank the UMMS MD/PhD program for ongoing support. This work was supported by a grant from the National Institute of General Medical Sciences (NIGMS) to AB.
Author contributionsAB supervised the project. CEF, ND, MT, JLL and AA carried out most of the experiments. KM, KB and YF provided essential reagents. CEF and AB discussed and interpreted the results and wrote the manuscript.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.